Study Links ChatGPT Use to Decline in Creativity and Memory Recall
A new preprint study from researchers at the Massachusetts Institute of Technology raises questions about the cognitive effects of using large language models like ChatGPT. The research, which has not yet been peer-reviewed, suggests that individuals who rely on AI to help with writing tasks may perform worse in measures of memory, creativity, and writing quality compared to those who complete the same tasks unaided.
The study adds to ongoing academic and public debates about the educational and psychological impact of generative AI tools, particularly among students and young adults. As AI systems become increasingly integrated into schoolwork and professional settings, researchers and educators are grappling with how they might influence critical thinking and original expression.
Though the sample size in this research is relatively small—just 54 participants—the findings align with broader concerns about “cognitive offloading,” a term used to describe the reliance on external systems to handle mental tasks that humans would otherwise do themselves.
Comparing Writing Approaches
Led by Nataliya Kosmya of MIT, the research team divided participants, aged between 18 and 39, into three groups. One group wrote essays using ChatGPT, another used Google Search for assistance, and the final group wrote without any external help. The study spanned four months and aimed to assess how each group performed across various cognitive and linguistic benchmarks.
According to the researchers, the group that used ChatGPT consistently underperformed compared to those who wrote unaided. The “Brain-only” group—those who did not use AI or search tools—scored higher across multiple areas, including essay structure, clarity, and originality. Participants in the ChatGPT group also struggled to recall content from essays they had written only minutes earlier.
The team reported diminished neural and linguistic engagement among ChatGPT users, suggesting that use of the AI tool may suppress deeper cognitive processing during writing tasks. While the researchers acknowledge that further study is needed, especially with a larger cohort and peer review, the early results add weight to growing interest in the long-term effects of AI-assisted work.
Caution and Context
The findings are not definitive but contribute to a larger body of evidence examining how AI might change not just how people work, but how they think. While AI tools like ChatGPT can improve efficiency, they may also promote passivity in tasks that traditionally require active learning, memory retention, and analytical thinking.
When asked about the study’s conclusions, OpenAI’s ChatGPT noted that such results are “interesting and plausible,” but cautioned that they should be interpreted carefully given the study’s limited scope. The model also suggested that cognitive offloading is not inherently negative, depending on the task and context.
Academic experts continue to debate the role of generative AI in education. Some argue that these tools should be embraced for their ability to support learners, especially those who struggle with reading and writing. Others caution that overreliance could lead to a decline in fundamental skills and deeper cognitive engagement. As AI tools become more accessible, the question of how and when to use them remains central to discussions in education, psychology, and technology policy.